Amazon has unveiled a new version of its own image generation model, the Titan Image Generator.
The mentioned virtual product is designed for Amazon Web Services (AWS) customers who use Bedrock’s generative artificial intelligence platform.
The new version of the specified model is called Titan Image Generator v2. This digital product has received several new capabilities. AWS principal developer advocate Channy Yun posted the relevant information on his blog.
As part of the interaction with Titan Image Generator v2, users can guide the images they generate using reference images, edit existing visuals, remove backgrounds, and generate variations of images. Channy Yun reported this.
According to the AWS principal developer advocate, the new version of the mentioned model can intelligently detect and segment multiple foreground objects. Channy Yun says that users will be able to generate color-conditioned images based on a color palette. It was also noted that consumers will be able to apply the image conditioning feature to shape their creations.
Titan Image Generator v2 supports image conditioning. If necessary, in this case, a reference image is used and focus is carried out on specific visual characteristics, including edges, object outlines, and structural elements. The model can also be fine-tuned using reference images, such as a product or company logo. This solution allows the generated images to maintain a consistent aesthetic.
AWS has not yet reported exactly what data is used in the Titan Image Generator training process. This Amazon unit noted in a comment to the media that in this case applying is a combination of proprietary and licensed arrays of information. It is worth noting that most vendors do not seek to disclose such details of their activities. The relevant behavioral practice is related to the fact that companies perceive training data as a competitive advantage and for this reason do not report such information, identifying it as something like a corporate secret. Moreover, the fact that vendors mostly refuse to disclose the mentioned details of their activities is explained by a kind of legal reason. Potentially, the company may face lawsuits if the arrays of information used to train models are disclosed. In this case, it means lawsuits related to intellectual property.
AWS implements an indemnification policy as an alternative to transparency. The relevant policy applies to customers in situations where Titan Image Generator v2 regurgitates a potentially copyrighted training example.
Amazon chief executive officer Andy Jassy, in the company’s earnings call for the second quarter of the current year, stated that he still evaluates the prospects of artificial intelligence technologies in an optimistic context of perception. In this case, it implied, among other things, digital products such as AWS Titan models. Andy Jassy continues to be optimistic, despite signs of doubt about the significant prospects for AI from the enterprise and the rising costs associated with training, fine-tuning, and serving models. He says that the development in the artificial intelligence industry will be fast. Also, according to him, in this case, the development process will be conducted from the get-go in the cloud.
As we have reported earlier, AWS Launches New GenAI-Powered Assistants.
